Part 3 of 4 · 2 min read
Exercises
Easy
Easy 1: Run it (30 minutes)
Run python -m pytest -q and python run_all.py in code/mini-pr (git must be installed).
Done when: you can explain each real pytest result in the table.
Easy 2: Read a contributing guide (1 hour)
Read the contributing guide of a project you use and list its rules on tests, style, commits and AI assistance.
Done when: you have a checklist of at least six rules.
Easy 3: Reproduce an issue (1 hour)
Pick an open bug in a project you use and reproduce it locally.
Done when: you have a failing command or test and a note of the version.
Medium
Medium 1: Review your own diff (2 hours)
Run check_diff on a branch of your own and fix or justify every flag.
Done when: the diff is smaller or each remaining flag is explained.
Medium 2: Write the description (2 hours)
Write a pull-request description for your capstone fix with issue link, why, what and how tested; run check_description.
Done when: there are no flags and a stranger could review it.
Medium 3: Add a rule (3 hours)
Add a rule for one more common review comment (for example a missing changelog entry) with tests.
Done when: you can show a diff that triggers it and one that does not.
Hard
Hard 1: A real contribution (10 hours)
Make a documentation, test or bug-fix contribution to a real agent-related open-source project and take it through review. (I did not run this, so your experience is the ground truth.)
Done when: the pull request is merged or closed with feedback you can summarise.
Hard 2: Break my checker (4 hours)
Produce three diffs that are bad but pass check_diff (for example a breaking rename inside a larger edit) and describe what check would catch them. (I did not run this.)
Done when: you have three examples and a proposed rule or review step for each.
Hard 3: Retrospective (3 hours)
After a real pull request, write what the maintainers asked for and compare it to what a checker could have found. (I did not run this.)
Done when: you list the share of comments a rule could have prevented.